14 Aug ESS IA Graphs: Chart Choices, Error Bars and Clear Labels
An ESS IA graph should help answer the research question. Choose it from the variables and the comparison you need, then check that labels, scale and data treatment are accurate. A trendline, equation, R² or error bars are useful only when justified; none belongs automatically on every graph.
Match the graph to the question
| Purpose | Useful options | Check |
|---|---|---|
| Relationship between two numerical variables | Scatter plot | Keep each paired observation together. Add a fitted model only if its form and assumptions suit the purpose. |
| Change through time | Time-series line or points | Use the actual dates or time intervals; connecting points implies a meaningful sequence. |
| Compare groups | Dot plot, boxplot or justified summary chart | Show the distribution or spread where possible rather than hiding every observation behind a mean bar. |
| Frequency distribution | Histogram or discrete frequency plot | Choose appropriate bins or count values and label frequency or density correctly. |
| Category proportions | Bar or stacked-bar chart; pie chart where suitable | Identify the denominator and categories. Check whether the plot makes comparisons clear. |
For matched before-and-after data, connected pairs or a plot of differences can show change more clearly than separate mean bars. A before-versus-after scatter plot can show association, but association alone does not describe the size of the change.
Use accurate labels and transparent scales
Label variables, categories and units where applicable. For example, dissolved oxygen concentration may be measured in mg L⁻¹; pH is dimensionless. Not every variable has an SI unit. Use precision consistent with the measurement, but do not force every axis tick to have the same number of significant figures as a raw observation.
Bars encode magnitude through length and normally need a zero baseline. Scatter plots and time series can use a restricted range when it is clearly labelled and does not misrepresent the variation. Zero is a valid temperature in °C and a possible pH value; it is incorrect to claim those scales cannot reach zero. A logarithmic axis must be identified and interpreted accordingly.
Use a legend or direct labels when more than one series needs identification. Distinguish series by symbols or line styles as well as colour, and keep text legible at the final size in the report.
Show spread or uncertainty appropriately
Repeated measurements can support an estimate of spread, but error bars are not mandatory on every figure. Showing individual observations may be more informative. State what the number of observations represents: repeat readings, independent samples or sites.
If using bars for uncertainty, identify their meaning and calculation. Standard deviation describes spread; standard error and confidence intervals address uncertainty in an estimate under their assumptions. A range is different again. None should be substituted for another merely because the software makes it easy.
Build the figure from checked data
Google Sheets
- Organise the original observations and any calculated summaries in clearly labelled columns.
- Select the relevant range and use Insert → Chart. Check the chart type and which columns are assigned as variables or series.
- Use Customise → Chart & axis titles for titles and axis labels. Axis-format settings control limits and appearance.
- Add a trendline through the series options only when a model is justified. Do not assume a linear fit is appropriate for most environmental data.
- Check the supported error-bar options against what you need to display.
Google’s official error-bar guidance describes constant, percentage and series-standard-deviation options for supported bar or line charts. It does not provide the custom per-point SD-column workflow claimed in some tutorials. A series-wide standard deviation is not the same as the separately calculated SD of each group. Use a suitable tool when the required representation is unsupported; do not change the scientific graph type just to obtain error bars.
Microsoft Excel
Choose the appropriate chart type, add axis titles and check variable assignments. In supported desktop versions, Chart Design → Add Chart Element → Error Bars → More Error Bar Options provides custom error amounts through Custom → Specify Value. Match each positive and negative value to the corresponding plotted point. Availability and menu names can differ by version.
Microsoft’s error-bar documentation explains the supported options. Check the selected cells and the meaning of the calculation; the software cannot decide whether SD, an interval or another measure is appropriate.
Check readability and export
Use strong contrast, distinguishable markers and a simple background. Avoid 3D effects that distort comparisons. Inspect the graph in greyscale and at the size it will occupy in the final PDF.
Keep an editable version and export or copy a clear figure using your application’s supported workflow. A PNG or suitable vector format can preserve sharp labels. There is no universal IB rule that every chart must be exported as a 300-DPI TIFF, and saving a workbook is different from exporting a chart image.
See Microsoft’s chart-as-picture guidance. After embedding, open the final report and check that labels, symbols and uncertainty remain readable; do not rely only on the spreadsheet view.
Place and explain essential figures
Keep graphs needed for the argument in the main report, close to the discussion. Number figures consistently, identify what is plotted and include relevant sample sizes and uncertainty definitions. Credit secondary data and explain your processing. An appendix may hold supplementary material but cannot replace the evidence needed to follow the argument.
The ESS report has a 3,000-word maximum for the course first assessed in 2026. Do not invent an additional universal page limit. Follow the guide’s detailed word-count rules and your school’s submission instructions.
Avoid misleading interpretation
R² is a model-fit measure and does not show whether a relationship is positive or negative. In a suitable simple regression, examine the slope and plotted pattern for direction. Correlation does not establish causation, and a visually striking trend does not by itself establish statistical significance.
A line of best fit does not automatically require its equation and R² printed inside the chart. Include and interpret model information where it helps answer the question, without cluttering the figure. Never select a model only to maximise R².
A caption pattern to adapt
“Figure [number]: [measured variable] across [sites or groups] during [period]. Points show [individual observations or calculated summary]; n = [independent units] per group. Error bars represent [defined measure], where shown. Data source: [your collection or cited dataset].”
Use only your actual values and methods. This is a text template, not a claim that downloadable Sheets or Excel files are supplied here. Do not copy invented regression equations, R² values or sample sizes into your investigation.
Final checks
- Does the graph answer a relevant part of the research question?
- Do plotted values match the source data and calculations?
- Are labels, units or category definitions accurate?
- Is the scale transparent, and is the type of graph justified?
- Are sample size and any spread or uncertainty defined?
- Is a fitted model justified and interpreted without causal overclaiming?
- Can the reader follow the figure in the final PDF?
Formatting supports communication; it does not guarantee marks. Use data-driven tools and check their output rather than generating an illustrative picture that merely looks like a graph. Follow school rules for tools and acknowledge assistance where required.
Discuss your graph choices
Marija Maletz is an ESS teacher, examiner and tutor, with 13 years of teaching experience and more than 10 years as an IB ESS examiner. Bring one graph, your question and a small example of the underlying data to discuss what you want it to show. The data treatment, interpretation and assessed writing remain your own.
Request a FREE 20-minute trial lesson to discuss needs and a suitable starting point. A full graph-by-graph review is not promised within the introduction.
No Comments